RevMD-Mem

Documentation Index

Fetch the complete documentation index at: https://docs.revilico.bio/llms.txt

Use this file to discover all available pages before exploring further.

​

Why Use This Engine?

In the documentation below, we will use Revilico’s RevMD-Mem engine to simulate a small molecule compound interacting with a lipid bilayer membrane, providing insight into passive membrane permeability, lipid partitioning, and drug-lipid interactions. These properties are directly relevant to ADME prediction: a compound that cannot permeate cellular membranes cannot reach intracellular targets regardless of binding affinity. RevMD-Mem quantifies the free energy cost of membrane crossing, enabling the prioritization of compounds with favorable permeability profiles before experimental testing.

​

Background

The lipid bilayer is a selectively permeable barrier composed of two leaflets of phospholipid molecules arranged with their hydrophilic headgroups facing aqueous solution and their hydrophobic acyl tails forming the membrane interior. Passive membrane permeation requires a compound to partition from the aqueous phase into the hydrophobic core, traverse the bilayer, and re-partition into the aqueous phase on the opposite side. The free energy profile along this permeation pathway, known as the potential of mean force (PMF), directly determines the permeation rate and membrane residence time. Molecular dynamics simulation of a compound within an explicit lipid bilayer provides an atomistically detailed view of drug-membrane interactions that continuum solubility models cannot capture. RevMD-Mem uses the GROMACS simulation engine with the CHARMM36 force field for lipids and the CGenFF (CHARMM General Force Field) for small molecules, generating PMF profiles and partition coefficients that predict passive permeability.

​

Simulation Pipeline

Membrane and System Construction A pre-equilibrated lipid bilayer patch (POPC by default, or DPPC for saturated membrane models) is used as the starting membrane geometry. The bilayer is oriented with the membrane normal along the z-axis. The compound of interest is initially placed in the aqueous phase at a specified distance from the bilayer surface. A periodic water box is added above and below both leaflets, and ions are added to 0.15 M NaCl. The resulting system contains approximately 128 to 256 lipid molecules, approximately 8,000 to 16,000 water molecules, and one or more drug molecules. Force Field Parameters Lipid parameters are taken from the CHARMM36 lipid force field, which has been extensively validated against experimental structural and thermodynamic data for biological membranes. Small molecule parameters are generated using the CGenFF parameterization tool, which assigns atom types and partial charges by analogy to known fragments in the CGenFF library. The same potential energy function as RevMD-Bind governs all interactions: $$E_{\text{total}} = E_{\text{bonds}} + E_{\text{angles}} + E_{\text{dihedrals}} + E_{\text{VDW}} + E_{\text{elec}}$$ Long-range electrostatics are handled with particle-mesh Ewald (PME) summation. A semi-isotropic pressure coupling (Parrinello-Rahman) is applied so that the membrane area and thickness can relax independently, which is essential for correct bilayer mechanics. Equilibration System equilibration follows the same NVT then NPT protocol as RevMD-Aqua, with heavy atoms initially restrained. The membrane-specific NPT equilibration uses semi-isotropic pressure coupling to allow the bilayer area per lipid to relax to its equilibrium value while maintaining the correct surface tension. Umbrella Sampling for PMF Calculation To compute the free energy profile along the membrane permeation pathway, the compound is driven through the bilayer using umbrella sampling. A series of simulation windows is generated by placing the compound at evenly spaced positions along the z-axis from the aqueous phase through the membrane center and out the other side. The spacing between windows is 0.1 to 0.2 nm, producing 30 to 50 windows covering the full bilayer thickness. In each window a harmonic restraining potential: $$V_{\text{umbrella}}(z) = \frac{1}{2} k_{\text{pull}} (z - z_0)^2$$ keeps the compound near its target z-position $z_0$ with a spring constant $k_{\text{pull}}$ of approximately 1000 kJ/mol/nm$^2$. Each window is simulated for 5 to 10 ns to sample the local free energy landscape. WHAM (Weighted Histogram Analysis Method) The biased distributions from all umbrella windows are combined using WHAM to reconstruct the unbiased potential of mean force: $$F(z) = -k_BT \ln \rho(z)$$ Where $\rho(z)$ is the unbiased probability density of the compound at position z along the bilayer normal. The PMF is computed iteratively until the free energy estimates converge. The resulting profile shows the free energy barriers to membrane entry, the partition free energy into the hydrophobic core (log $P_{\text{mem/water}}$), and the overall free energy cost of traversing the bilayer.

​

Analysis Outputs

Potential of Mean Force Profile The PMF plot shows the free energy in kcal/mol as a function of position along the membrane normal. Key features include the barrier height at the membrane-water interface (related to initial partitioning), the depth of the free energy minimum in the hydrophobic core (the membrane affinity), and the overall permeation barrier from one aqueous phase to the other. A low, broad minimum with small interfacial barriers characterizes lipophilic compounds that partition efficiently into membranes. A high central barrier with a deep minimum characterizes “membrane-trapped” compounds that partition into but do not easily permeate the bilayer. Log Kp (Membrane-Water Partition Coefficient) The free energy minimum relative to the aqueous bulk gives the membrane-water partition coefficient: $$\log K_p = \frac{-\Delta G_{\text{min}}}{k_BT \ln 10}$$ Permeability Coefficient The overall passive permeability is estimated from the PMF and diffusion coefficient profiles using the inhomogeneous solubility-diffusion model: $$P = \left[ \int_{-d/2}^{d/2} \frac{e^{F(z)/k_BT}}{D(z)} \, dz \right]^{-1}$$ Where $D(z)$ is the position-dependent diffusion coefficient estimated from the autocorrelation of coordinate fluctuations in each umbrella window. Lipid Interaction Analysis The compound’s preferential interactions with specific lipid headgroup atoms and acyl chain carbons are quantified through radial distribution functions and contact analysis, identifying whether the compound intercalates into the hydrophobic core or associates preferentially with specific lipid regions.

​

Running the Engine

​

Inputs

​

Outputs

• PMF profile: Free energy along membrane normal with statistical uncertainty
• Log Kp: Membrane-water partition coefficient
• Permeability coefficient: Estimated passive permeability in cm/s
• Bilayer thickness and area per lipid: Structural properties during simulation
• Compound tilt and orientation plots: Preferred molecular orientation within the bilayer
• Lipid contact analysis: Preferential interactions with specific lipid components
• Convergence analysis: PMF bootstrap uncertainty and per-window sampling quality